import tkinter as tk
root = tk.Tk()

def bad(): print(a, 'bad')
def good(): print('good')

root.after(1000, bad)
root.after(1500, good)
root.mainloop()


# Correctly gives this traceback and output:

Exception in Tkinter callback
Traceback (most recent call last):
  File "C:\Programs\Python37\lib\tkinter\__init__.py", line 1699, in __call__
    return self.func(*args)
  File "C:\Programs\Python37\lib\tkinter\__init__.py", line 745, in callit
    func(*args)
  File "F:\Python\a\tem2.py", line 13, in bad
    def bad(): print(a, 'bad')
NameError: name 'a' is not defined
good
>>> 
====================================

Remove or comment-out the blocking 'root.mainloop()' call and run the result from an IDLE editor.  The callbacks are still called because after user code is run, run.py calls tcl.update in a loop nominally 20 x per second.  This allows developers to interact with a 'live' gui by entering statements in the shell at '>>>' prompts.  The result is this.
-------------
>>> Exception in Tkinter callback
Traceback (most recent call last):
  File "C:\Programs\Python37\lib\idlelib\run.py", line 137, in main
    seq, request = rpc.request_queue.get(block=True, timeout=0.05)
  File "C:\Programs\Python37\lib\queue.py", line 169, in get
    raise Empty
queue.Empty

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
<The NameError traceback and message, as before.>
--------------

The relevant code in run.py was written before callback chaining.

            try:
                seq, request = rpc.request_queue.get(block=True, timeout=0.05)
            except queue.Empty:
                handle_tk_events()
                continue

Experiments with normal exceptions in a shell suggest that wrapping handle_tk_events in try:except and re-raising any exception 'from None' should work.
                try:
                    handle_tk_events()
                except BaseException as e:
                    raise e from None

However, it appears that callback errors resulting from handle_tk_events() are not being caught here.  (print('message') and 1/0 before 'raise' have no visible effect.)  I will investigate more later.

try:
                seq, request = rpc.request_queue.get(block=True, timeout=0.05)
            except queue.Empty:
                request = None
            if request is None:
                handle_tk_events()
                continue

Thanks.  I am still puzzled why the nested within nested try-except did not work as I expected, but I care more about fixing this. The result based on your suggestion is better (to read, I think) than extra nesting that did work.

New changeset 1e2fcac4972530aa2c963d7e4011021df5ba866e by Terry Jan Reedy in branch 'master':
bpo-32207: Improve tk event exception tracebacks in IDLE. (#4703)
https://github.com/python/cpython/commit/1e2fcac4972530aa2c963d7e4011021df5ba866e

New changeset 9da33c82124f27eb58ba4cf145675fe7a1035744 by Terry Jan Reedy (Miss Islington (bot)) in branch '3.6':
bpo-32207: Improve tk event exception tracebacks in IDLE. (GH-4703) (#4705)
https://github.com/python/cpython/commit/9da33c82124f27eb58ba4cf145675fe7a1035744